LLaVA-CoT Shows How to Achieve Structured, Autonomous Reasoning in Vision Language Models https://t.co/PQyW8vpbQg by Sergio De Simone

Ivan Fioravanti ᯅ@ivanfioravanti

2 years ago

A preview of MMLU PRO scores in Computer Science for Qwen2.5-72B-Instruct using Apple MLX: • 4bit: 78.78% (4h 40m) - 323/410 correct answers • 3bit: 71.46% (3h 38m) - 293/410 correct answers In brackets testing time, yes you read it well: 8 hours to test 2 quantizations 👀… https://t.co/YUlmJp9Js6

Rohan Paul@rohanpaul_ai

2 years ago

LLaVA-o1 teaches machines to think step-by-step like humans when analyzing images. LLaVA-o1 introduces a novel approach to enhance Vision Language Models (VLMs) by implementing structured, multi-stage reasoning. This paper tackles the challenge of systematic reasoning in visual… https://t.co/jrSutB9Dat